Mammal dung–dung beetle trophic networks: an improved method based on gut-content DNA

Beetle collection

We collected dung beetles from a temperate forest in Germany (49°51′54.19″N, 8°41′28.50″E) and a lowland tropical rainforest, as well as five pastures in Ecuador (0°30′20.52″N, 79°10′31.95″W Ecuador). Collections were made in September 2019 (Germany) and from January 2019 to June 2019 (Ecuador). Beetle collections were made under the Escuela Politecnica Nacional, Contrato Marco MAE-DNB-CM-2016-0068, and transported to Germany under beetle export authorization number 62-2019-EXP-CM-FAUDNBIMA. Beetles were collected using pitfall traps, baited with cow dung in Germany (seven pitfall traps, 41 beetles), and human dung in Ecuador (eight pitfall traps, 94 beetles). Beetles were also collected opportunistically by hand in Ecuador (N = 41).

Both cow and human dung are common attractants for dung beetle pitfall trapping (Hanski & Cambefort, 1991; Whipple & Hoback, 2012; Marsh et al., 2013; Frank et al., 2018a; Frank et al., 2018b). However, in preliminary trials, we realized that, in the Ecuadorian population, cow was less effective in attracting dung beetles than human dung. The latter attracted a higher number and diversity of dung beetles, so we decided to use human dung in Ecuador. We did not see a substantial difference between bait types in Germany and, therefore, decided to use cow dung as it is easier to handle. In all experiments, we euthanized beetle specimens quickly using either absolute ethanol or freezing to reduce the suffering of the specimens. Disposable nitrile gloves (VWR) were worn for all dung manipulations for personal protection, especially as human dung was used as a bait. After the baiting periods, human dung was disposed of in the same hole created for the pitfall trap following the Leave No Trace Principles for human waste disposal (www.lnt.org).

To minimize the contamination of beetle guts with cow or human dung, direct physical contact from dung beetles to baits was minimized. The pitfall traps consisted of a plastic cup leveled with the soil, a rain cover, and an overhanging tea bag with bait inside, mostly prohibiting unintended bait consumption (Frank et al., 2018a). Pitfall traps were installed 24 h before collection. When pitfall traps were collected, only dung beetles were collected. All other animals were released. After beetle collection in the field, specimens were either preserved in ethanol and then frozen (Ecuador) or simply frozen (Germany). Our field site in Ecuador is subject to occasional power outages and thus requires a second level of DNA preservation. Dung beetles were then identified morphologically by KMP using the latest species keys (Edmonds, 2000; Solís & Kohlmann, 2002; Edmonds & Zídek, 2004; Edmonds & Zídek, 2012; Vaz-De-Mello et al., 2011; Chamorro et al., 2018; Nunes, Nunes & Vaz-de Mello, 2018), and the reference collections of the Pontificia Universidad Católica (Ecuador) and the Escuela Superior Politécnica del Litoral (Ecuador).

Beetle dissection

After frequently amplifying human DNA in preliminary experiments before this study began, we developed a washing protocol in sterile conditions to minimize human DNA contamination, which markedly reduced the amplification of contaminants such as human DNA and prevented cross-contamination of samples. First, surfaces and tools were UV sterilized. Beetles were placed singly in distilled water and shaken for 30 s to remove dirt and external DNA from the outside. We transferred the beetles in a 2% NaClO (chlorine) solution under a fume hood where the air was constantly UV sterilized. The chlorine solution was washed off the beetles with 70% ethanol. Then, beetles were transferred to a dissection tray and dissected using various dissection tools. Tools were flamed, washed in 2% NaClO, rinsed in 70% EtOH, and then flamed again and cooled before each dissection. Cleaning solutions were changed every two beetles to prevent cross contamination and reduce waste liquid byproducts. After every beetle, the dissection tray was washed with 2% NaClO solution and 70% EtOH. We dissected 177 beetles, 135 from Ecuador and 42 from Germany.

For large beetles (body length > 1 cm), the digestive tract was removed and placed in the DNA extraction buffer of a Qiagen Blood and Tissue Kit (Qiagen, Hilden, Germany). For smaller beetles (body length < 1 cm), the entire abdomen was placed in the same DNA extraction buffer without further dissection. We used the entire abdomen for small specimens to prevent possible contamination. Their small size made it much harder to dissect, and as our first attempts at sequencing this way worked, we kept using the method. For a subset of 51 large beetles, we visually categorized gut fullness in the following way: (1) full—more than half the length of the intestines is full; (2) half-full—half or less than half the length of the intestine is full; (3) empty—there is no visible content in the intestine (Fig. S1).

DNA extraction and amplification

DNA was further purified using the Bio-RAD Micro Bio-Spin Columns P-30 Tris following the manufacturer’s instructions. This results in 75 µl of purified DNA, which can be used for PCR. We then ran PCRs using the Qiagen multiplex kit (with 1.5 µl of molecular grade water), 5 µl of Qiagen multiplex PCR Master Mix, then 0.5 µl of 10 µM for the forward and reverse primers, and with the addition of 0.5 µl of bovine serum albumin (BSA) to counteract remaining PCR inhibitors and 2 µl of template DNA. The PCRs started with an initial activation period of 95 °C for 15 min, followed by 35 cycles with the following settings: denaturation at 94 °C for 30 s, the annealing temperature of 65 °C for 90 s; extension at 72 °C for one minute. A final elongation step was performed at 72 °C for 10 min. PCRs were repeated up to three times per sample. For all 177 beetles, we used vertebrate-specific PCR primers to amplify portions of the mitochondrially encoded 12S rDNA (Ushio et al., 2017). For the Ecuadorian beetles (N = 135 beetles), we additionally used the mammal-specific primers 16smama1 (forward) and 16smama2 (reverse) (Taylor, 1996). For a subset of samples (N = 121 DNA extractions), we measured the DNA concentration using a ThermoScientific Nanodrop Lite Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Of 177 beetles, 128 were amplified and sent to sequencing.

Proof of concept study

As a proof of concept, we offered a defined diet to Anoplotrupes stercorosus dung beetles consisting of a variety of mammals and amplified mammal DNA from the beetles’ guts. We then amplified the mammal DNA from the beetles’ guts. All 30 beetles were initially fed apples for five days to clear their guts. We then fed them with 10 grams of the following six dung types: tapir (Tapirus terrestris), fennec fox (Vulpes zerda), otter (Aonyx cinerea), porcupine (Hystrix cristata), macaque (Macaca nigra), and cow (Bos taurus). Beetles were fed for 24 h to ensure they had enough time to consume dung. Dung types were fed to five beetles each. Dung beetles were then euthanized in the freezer. The dung was contributed by a local zoo (Vivarium Darmstadt) and a local farm. We then applied our method as described above to identify dung beetle gut content. The amplified 12s rDNA fragment was then compared with a reference library using the Basic Local Alignment Search Tool (BLAST) to verify if it matched the consumed mammal dung (Altschul et al., 1990). Please note that we only used one species and a relatively low sample size for this proof-of-concept experiment, so caution should be taken in extrapolating the results to other dung beetle species.

Gut content fullness of beetles in pitfall traps after 8 h, 24 h, and 48 h

Pitfall trapping means that live beetles might empty their guts in the trap before being collected. At our German field site, we experimentally assessed how many beetles would still have full guts after pre-defined time periods. To approximate realistic gut fullness in pitfall traps, we set out 15 traps in the forest. The traps were baited with cow dung and randomly assigned to three different groups: five traps were emptied after 8 h, five traps were emptied after 24 h, and five traps were emptied after 48 h. Only A. stercorosus beetles were collected and immediately placed in 70% ethanol. In this proof-of-concept study, we solely used A. stercorosus to assess the level of gut fullness. Due to its larger size compared to Aphodius sticticus, the gut is easier to dissect reliably. In addition, they are also more readily available in Germany, and it is easier to identify the species in the field. The smaller Aphodius beetles are much harder to reliably identify. Beetle abdomens were then dissected to record gut fullness (at least 50% of the gut filled) or less than 50% filled. Data were analyzed with a Pearson’s chi-squared test.

Gut content after 48 h of starvation

A. stercorosus beetles (N = 20) were collected from the forest, transferred alive into different enclosures, fed cow dung for five days, and then starved for two days to survey the proportion of empty guts. After 48 h, the beetles were frozen and then dissected. Contents of guts (empty, not empty, full) were recorded together (Fig. S1). As a control, 20 beetles with ad libitum access to cow dung were frozen simultaneously to survey how many beetles had an empty gut when provided a continuous supply of food. Data were analyzed with a Person’s chi-squared test.

DNA sequence processing

For gut DNA analysis, we dissected 177 beetles from the two field sites belonging to 10 dung beetle species. Eight of the species were from Ecuador: Deltochilum sp.(8), Oxysternon conspicillatum (53), Canthon angustatus (58), Onthophagus sp. (8), Dichiotomius sp. (1), Canthidium sp.(3), Sulcophanaeus notis (1), and Scybalocanthon trimaculatus (2), as well as one unidentified species; two species were from Germany: A. stercorosus (19), and A. sticticus (23). PCR products of successful DNA amplifications, verified by gel electrophoresis and staining with ROTI® GelStain, were sent for Sanger sequencing to Macrogen Europe. Forward and reverse directions were sequenced for each amplicon. Post-processing was done using Codon Code Aligner 10.0.2 on macOS High Sierra. Low-quality base pairs (base pairs with a quality score lower than Phred 20) were clipped from the ends of the sequences. A Phred 20 quality score corresponds to 99% accuracy in a base call. At this stage, low-quality sequences were discarded (sequences with 50 or more base pairs with a score lower than Phred20) or sequences with a length of less than 50 base pairs. The resulting sequence length and quality scores were then recorded. Sequences were aligned, if possible, to create a consensus sequence. The resulting consensus sequences were then matched to reference sequences using the NCBI MegaBLAST search (Morgulis et al., 2008). We accepted the best match as our ID if the match was >90%, which is commonly used for 12s rDNA and 16s rDNA short sequences at the genus level for mammals (Hoffmann et al., 2017; Kocher et al., 2017; Drinkwater et al., 2019; Saenz-Agudelo et al., 2022). The best species match was recorded, along with the accession ID, percent identity, max score, and bit score. Mismatches between references and query sequences were often found in base pairs with a low-quality score. Lower matches might have partly arisen from DNA degradation as we analyzed DNA within digestive tracts. Due to this constraint and previously established protocols (Hoffmann et al., 2017; Kocher et al., 2017; Drinkwater et al., 2019; Saenz-Agudelo et al., 2022), we decided to identify the mammal species only to the level of the genus when there where multiple species within a genus otherwise we used the species name.

Data analysis

For the pitfall trap dataset, we ran a logistic regression with positive electrophoresis results as the dependent variable and gut fullness, collection method, PCR primers, DNA concentration, and beetle species as predictor variables. We then performed a dominance analysis (Budescu, 1993) to determine which of the predictor variables were the most important predictors of DNA amplification. A dominance analysis compares predictor variables in a pairwise fashion across all the subset models and generates a predictor hierarchy or the importance (dominance) of each predictor value. Higher values have a greater predictive power (Azen & Budescu, 2006; Lee & Dahinten, 2021).

We also constructed bipartite food networks (mammal dung–dung beetle) for each site to investigate the specificity of dung choice. For this, we counted the number of beetle individuals containing DNA of a given mammal genus. Our Ecuador data set includes 41 beetles whose gut content DNA successfully amplified and from which we successfully obtained sequence matches out of a total of 135 dissected beetles. The German network includes data from 21 beetles of the total 42 dissected beetles. Our ‘interaction frequency’ was the number of beetle individuals per species that contained a specific mammal DNA sequence. We calculated network specialization using the H₂′ statistic for each plot (Blüthgen, Menzel & Blüthgen, 2006) as in Frank et al. (2018b). We compared H₂′ values to randomized networks using the Patefield null model (see Blüthgen, Menzel & Blüthgen, 2006). We excluded sequences matching the bait (human in Ecuador and cow in Germany). Beetle species with no measured trophic interactions with mammal dung were dropped from the network analysis. We included humans as dung beetle interaction partners in our German network for two reasons. First, we observed a fair amount of human excrement at the German study site. Second, we showed that the newly developed washing protocol drastically decreased the detection of human contamination in the samples. Although contamination with human DNA cannot be entirely excluded, we consider it likely that most of the detected interactions with human dung at the German study site are accurate. When displaying the networks, we use species names if only one species of the genus is present in the study site, or we use spp. to indicate that more than one species is possible; however, we caution that using species distribution information in this way could make investigators overlook species with expanding ranges, or closely related genera could be misidentified. Finally, to estimate the diversity of mammal dung consumed by beetle species within each habitat type, we calculated the Shannon index for each beetle species and their associated mammal diversity.

Proof of concept

We offered six dung types to a single dung beetle species. We verified the mammal genera for all six test dung types by matching DNA barcodes from beetle gut contents. The best DNA barcode match corresponded to the respective species for tapir, otter, porcupine, and cow (the top 10 best matches are the expected species), but macaque and fox have mixed species in the top 10 best matches. The 16s rDNA or 12s rDNA fragments for mammal identification are often used for genus level identification because this increases the accuracy of identification (Hoffmann et al., 2017; Kocher et al., 2017; Drinkwater et al., 2019; Saenz-Agudelo et al., 2022). Sequence quality was high, with a Phred score of 20 or higher from all base pairs. Data are available in the Supplemental Information.

Pitfall trap gut content fullness after defined hours

Among all beetles from traps emptied after eight hours, 75% had full guts (N = 15 total beetles, four beetles with empty guts and 11 beetles with full guts). After 24 h, 61% of the beetles had full guts (N = 44 total beetles, 17 beetles with empty guts and 27 beetles full guts), and 58% had full guts after 48 h (N = 24 total beetles, ten beetles with empty guts, and 14 beetles with full guts). Despite a decreasing proportion of full guts over sampling time, the three time points did not differ significantly in their proportion of full guts (Pearson’s chi-squared test: Chi² = 0.95, df = 2, p = 0.621).

Gut fullness after 48 h of starvation

In a trial testing how many beetles could be expected to retain full guts after 48 h, we found that among the 20 beetles included in the starvation treatment, half the beetles (N = 10) had empty guts, nine guts were half full, and only one gut was full. Among the 20 beetles where food was available ad libitum for 48 h, five beetles had empty guts, five beetles had half full guts, and ten beetles had full guts. The proportion of full guts is thus lower in the starvation treatment than in the fed treatment. Accordingly, the two treatments differed (Pearson’s chi-squared test: Chi² = 10.17, df = 2, p = 0.006).

Field collected dung beetles for gut content identification

Networks

With the dung beetles from Ecuador, we used both 16s and 12s rDNA primers. The network for the 16s rDNA primers is much less complete than the network generated using the 12s rDNA primers (Figs. S2 and S3). The 16s rDNA Ecuador network has an H₂′ = 1, and the 12s rDNA network has an H₂′ = 0.66. Both are significantly different than the null model (both p < 0.001). However, after the data from both primers were combined, the level of network specificity dropped to H_2′ = 0.49, also significantly different from the null model (p < 0.001) (Fig. 1). The mammal dung–dung beetle network from Germany showed a similar moderate degree of specificity (H_2′ = 0.41, p = 0.005; Fig. 2). We detected dung of 14 mammal genera from 10 dung beetle species (Tables S2, S3 and S4). These included top predators (e.g., P. onca (jaguar)), herbivores (e.g., Capreolus capreolus (deer)), and omnivores (e.g., Caluromys sp. (opossum)) (Tables S5 and Table S6). Per beetle species, we detected between one and six mammal genera (Tables 3 and 4).